Pien versus Pim

For my corpus, I would like to compare my Spotify Wrapped playlist of 2019 to the one of my boyfriend, Pim. Our taste in music is very different and those lists are representing this difference.

I think it’s interesting to compare these two playlists. I’m more into soft pop music, singer-/songwriters, also a little classical music, film music, etc. And Pim is more into Drum & Bass, Rap, Hiphop, etc. But we appreciate each others taste of music increasingly since we’ve been together. For a research question I would think of how different our music tastes really are. At first glance, they are very different. But maybe there are some underlying similarities. And what those differenses and similarities mean.

The meaning of music is also different. Pim doesn’t like to really listen to en think about lyrics of songs. He just want to feel it physically, especially the beat, and dance or focus. I really love to feel music emotionally and think about lyrics. For me is music a way to express feelings.

I’m going to look at the similarities and differences in our playlists. First I will compare the whole playlists. By doing that, I will get a clear overview of my corpus. After that, I am going to take a deeper look by comparing specific songs of our playlists. I will be doing this for chroma and timbre, key and tempo.

Pim is dancing a lot more


I have some first findings, measured by Spotify, that are interesting: The music that Pim listens to is more danceable (M = 0.72, SD = 0.13) than the music I listen to (M = 0.48, SD = 0.19). Also there is a difference in energy. Pim’s playlist is way more energetic (M = 0.68, SD = 0.18) than mine (M = 0.34, SD = 0.25). If we look at the loudness category, Pim’s playlist is louder (M = -6.72, SD = 2.41) than mine (M = -14.1, SD = 8.79). In the figure below, we can see the differences that I discussed above. In Pim’s figure, almost all of the points/dots are on the top right. This means, that the the energy and danceability is very high. The loudness is indicated with the size of the points/dots. In my figure, the points/dots are more spread out. But almost all the points/dots are very small. So in the size (loudness), Pim is more diverse. And my figure is more stable. Which is a little strange because my standard deviation is a lot bigger than Pim’s. To make this more clear, I made a figure of the loudness.

To understand the difference in Loudness, I made a nice figure.


In the figure we see that Pim is more specific and I am more diverse and stable. This makes more sense. My standard deviation of loudness is large because I listen to more different types of music. A reason for this could be that I have to listen a lot of music for my study, musicology. And this is especially classical music. Therefore, I will look, from now on, more at the part where there are more similarities and leave the bottom left side of my playlist as kind of outliers. I will do this because otherwise the differences become really big and not very useful.

Our similarities!

This table shows us a summery of all the Spotify features. In the previous tab I discussed the difference in Danceability, Energy and Loudness. If we look at the Acousticness, there is also a quit large difference.

But there are also similarities like the Keys of our playlists. If we look at the results of Pim in this category, it would mean in the pitch class: F/F#. My results for Key would mean in pitch class: E/F.

Also the modes are close to each other. In the table we see that Pim en I both listen to major, minor and everything in between.

Apparently, Pim en I are both not really into cheerful music. We score pretty low on Valence, especially I am. But I didn’t expect that Pim would be pretty close.

The results of speechiness seems a little extreme to me. Despite the fact that Pim listens to rap, the results are very low. I would expect that the results of the speechiness in Pim’s playlist would be between 0.33 and 0.66. I don’t think I have to exclude this because it might be an interesting thing to investigate more.


Danceability Mean SD
Pim 0.72 0.13
Pien 0.48 0.19
Energy Mean SD
Pim 0.68 0.18
Pien 0.34 0.25
Loudness Mean SD
Pim -6.72 2.41
Pien -14.1 8.79
Acousticness Mean SD
Pim 0.13 0.19
Pien 0.63 0.36
Key Mean SD
Pim 5.43 3.74
Pien 4.54 3.35
Mode Mean SD
Pim 0.53 0.50
Pien 0.57 0.49
Speechiness Mean SD
Pim 0.14 0.11
Pien 0.06 0.06
Valence Mean SD
Pim 0.40 0.24
Pien 0.26 0.19
Liveness Mean SD
Pim 0.18 0.12
Pien 0.16 0.14
Tempo Mean SD
Pim 129 30.8
Pien 113 32.2

Apocalypse and Highest In The Room in a chromagram


There are only two songs that appear in both our playlists. Those songs are Apocalypse by Cigarettes After Sex and Highest In The Room by Travis Scott. These songs are very very different. That’s why I have put them together in a chromagram.

The Timbre and Chroma in Bad Guy and Hunnybee.

For the timbre and chroma I compare the two most danceable songs in our playlists. In my playlist is ‘Bury a Friend’ by Billie Eilish the most danceable song. In the playlist of Pim, ‘Hunnybee’ by Unkown Mortal Orchestra is the most danceable song. I chose these two songs because I think it would be interesting to look if these songs have similarities.

The grams of Bad Guy shows us that there is a structure in the song. And that changes at almost the end of the song. The structure stops. But according to the grams, the beat is considered to be almost the same as before (the purple blocks). Around 150, it turns into a little lighter purple but it’s almost the same.

In the grams of Hunnybee, there is not really a clear structure. There is a beat that continues trough the whole song. That is why the blocks are almost the same color, except the black one. The differences are in the lyrics of the couplets and refrains. Just after 200, something happens. Here is a guitarsolo and after that, there is again the refrain.

So these songs don’t have a lot similarities. They both have a deviation in their structure. But their structures are really different. What they do have in common is that in both songs, there is a beat that continues almost the whole songs. So I think that’s why they both score high on danceability.

Lions Don’t Cry and Aguella in keygrams


Here I made two keygrams of the songs in our playlists with the most energy. In my playlist the most energetic song is Lions Don’t Cry by Tim Akkerman and has an energy of 0.93. In Pim’s playlist the song with the most energy is Aguella - Original Mix by Sllash & Doppe and has a energy of 0.98. The keygrams are looking very different. Let’s take a look.

The keygram of Lions Don’t Cry is very clear. It says that almost the whole song Db major and G major are present. Also E minor occurs a lot. There are also a few other keys that are very dark in the gram. I can compare these with the information that Chordify gives about this song. According to Chordify this song is in E minor. Here is the G major right because that is the III. The Db major is a little bit strange. Because in E minor it can be a C#, that is the sixth increased (verhoogde zesde trap). But in Chordify there is a C major and D major, not something in between. But in the E minor key, it is possible. Also in the keygram, the D# minor is really dark. This is the vii. It is not present as chord in the song but for the key it is really logical. One thing that is really noticable is the B. According to Chordify B major and B minor are both present in this song. The B minor more often than the majeur chord. The B major would be logical because that is the fifth/V and this chord should be major. So why the B minor is more common, is a little but strange. Also both the B major and B minor are not really visible in the keygram. So beside the B minor/majeur thing, the key of the song (E minor) is very clear in both the keygram and Chordify.

The keygram of Aguella is not so clear. I think that this is a hard song to analyze because it is electronic music. In the keygram the B minor and major, F# minor and Ab major are really dark. According to Chordify, this song is in G# minor. The Ab major can be a confusion with G# because that is the same tone. And in chordify there is sometimes a G# major present. The G# minor is in the keygram quite light and that is really strange because the G# minor chords appears a lot in the song. Also according to Chordify, the chord D# minor appears a lot. But in the key G# minor, D# would be the fifth/V and that should be a major chord. But the key of this song (G# minor) is clear.

But of course these are keygram and both of them are showing the keys clearly. They don’t really show the chords in the songs. The songs don’t have a lot in common, only that they are both energetic. I think that the keygram of Lions don’t cry is more clear because that song is made with recognizable instruments and Aguella is really electronic and maybe harder to recognize.

Highest In The Room and Apocalypse are really different in tempo.


For the tempograms, I picked the only two songs that appear in both our playlists. Because these two songs seems to be very different, I chose to work these two out in tempograms. And maybe I can come to a conclusion why these two particular songs are the only two that are in both of our playlists.

You can see in the tempograms, that Highest In The Room by Travis Scott scores really high in tempo. It is almost the whole song about 155 BPM and sometimes it even reaches the 160 BPM.

In the tempogram of Apocalypse by Cigarettes After Sex that the this song is really slow. It is about 90 BPM. Sometimes it is almost 100 BPM and sometimes it is even on the lowest point of the tempogram, 80 BPM.

So I can conclude here that the two songs differs a lot in tempo. It is almost the exact opposite. Still we both really like these songs. Maybe our tastes in genres are not that different, only the levels in the genres.

Classification

          Truth
Prediction Pien Pim
      Pien   16   4
      Pim     4  16

# A tibble: 3 x 3
  .metric  .estimator .estimate
  <chr>    <chr>          <dbl>
1 accuracy binary           0.8
2 kap      binary           0.6
3 j_index  binary           0.6
# A tibble: 3 x 3
  .metric  .estimator .estimate
  <chr>    <chr>          <dbl>
1 accuracy binary         0.825
2 kap      binary         0.650
3 j_index  binary         0.650
$Pien
35 x 1 sparse Matrix of class "dgCMatrix"
                          1
(Intercept)       0.8711269
danceability     -0.8381242
energy            .        
loudness          .        
speechiness      -2.0432858
acousticness      1.2514814
instrumentalness  .        
liveness          2.5116214
valence           2.8206431
tempo             .        
duration          .        
C                 0.9419411
`C#|Db`          -0.1686037
D                -1.0204213
`D#|Eb`          -0.3507055
E                 2.0816303
F                 .        
`F#|Gb`          -0.5156069
G                 .        
`G#|Ab`          -0.3276726
A                 .        
`A#|Bb`           .        
B                 .        
c01               .        
c02              -0.5909089
c03               .        
c04              -1.4437086
c05              -2.4304637
c06               .        
c07               .        
c08               0.3024615
c09               2.0213563
c10               .        
c11               1.8719799
c12               .        

$Pim
35 x 1 sparse Matrix of class "dgCMatrix"
                          1
(Intercept)      -0.8711269
danceability      0.8381242
energy            .        
loudness          .        
speechiness       2.0432858
acousticness     -1.2514814
instrumentalness  .        
liveness         -2.5116214
valence          -2.8206431
tempo             .        
duration          .        
C                -0.9419411
`C#|Db`           0.1686037
D                 1.0204213
`D#|Eb`           0.3507055
E                -2.0816303
F                 .        
`F#|Gb`           0.5156069
G                 .        
`G#|Ab`           0.3276726
A                 .        
`A#|Bb`           .        
B                 .        
c01               .        
c02               0.5909089
c03               .        
c04               1.4437086
c05               2.4304637
c06               .        
c07               .        
c08              -0.3024615
c09              -2.0213563
c10               .        
c11              -1.8719799
c12               .        
# A tibble: 3 x 3
  .metric  .estimator .estimate
  <chr>    <chr>          <dbl>
1 accuracy binary         0.725
2 kap      binary         0.450
3 j_index  binary         0.450

Call:
C5.0.default(x = x, y = y, trials = 1, control = C50::C5.0Control(minCases =
 2, sample = 0))


C5.0 [Release 2.07 GPL Edition]     Thu Mar 26 17:46:23 2020
-------------------------------

Class specified by attribute `outcome'

Read 40 cases (35 attributes) from undefined.data

Decision tree:

acousticness > 0.08576353: Pien (15)
acousticness <= 0.08576353:
:...`C#\|Db` > -0.09054911: Pim (14)
    `C#\|Db` <= -0.09054911:
    :...c05 <= -0.09163761: Pien (4)
        c05 > -0.09163761: Pim (7/1)


Evaluation on training data (40 cases):

        Decision Tree   
      ----------------  
      Size      Errors  

         4    1( 2.5%)   <<


       (a)   (b)    <-classified as
      ----  ----
        19     1    (a): class Pien
              20    (b): class Pim


    Attribute usage:

    100.00% acousticness
     62.50% `C#\|Db`
     27.50% c05


Time: 0.0 secs
# A tibble: 3 x 3
  .metric  .estimator .estimate
  <chr>    <chr>          <dbl>
1 accuracy binary         0.775
2 kap      binary         0.55 
3 j_index  binary         0.550

# A tibble: 3 x 3
  .metric  .estimator .estimate
  <chr>    <chr>          <dbl>
1 accuracy binary          0.75
2 kap      binary          0.5 
3 j_index  binary          0.5 

I made a confusion matrix of my our playlists. You can see that the differences are pretty clear and that the predictions are about 75% correct. Butyou also can see here, that Pim is clearer than I am. It varies only two (or one) songs but the prediction that a song belongs in Pim’s playlist when it really belongs to mine, occur more than the prediction that a song belongs in my playlist when in fact it belongs in Pim’s. So I would conclude out this figure that I’ more flexible than Pim is.

You can also see this in the Timbre Components plot. The songs that belongs in my playlist, are more spread out over the whole plot. The songs of Pim’s playlist, are very close to each other. Also there is more variation in liveness in my playlist.

Conclusion

For now, I can conclude that my boyfriend and I really have different tastes in music genres. But sometimes there are a few similarities. I think that it would be really interesting to look in a few years, what the differences are then. It is a pity that I don’t have these wrapped up lists from the other years that we’ve been together. Because it would be interesting to compare all those lists.

I think that I listen to a lot more different genres than Pim does. I’ve focused in this portfolio on the most danceable and energetic songs, and the two songs that appear in both our playlists. I haven’t focused on the slowest songs because those diffences are really big.

In the results, we’ve seen that Pim’s playlist is more danceable and mine is more acoustic. I’ve also discussed our similarities and we both listen to major and minor but not really cheerful music. The mean in key of my playlist is E/F. And de mean in key in the playlist of Pim is F/F#. If we compare this to the keygrams I made, it is right. The keygram of Lions Don’t Cry from my playlist is in E minor. And the keygram of Aguella from Pim’s playlist is in G# minor. My most energetic song is a little bit closer to my mean than Pim’s most energetic song.

If we look at the tempograms and compare these with the means of tempo, we see also something interesting. The mean of tempo in my playlist is 113. In Pim’s playlist the mean of tempo is 129. The two songs that I put in a tempogram, are really different. Their tempi is also really apart. The song Highest In The Room is for both our playlist really fast (155 - 160 BPM). The other song, Apocalypse, is really slow compared to our means (80 - 100 BPM).

What would be interesting is to look at the results from this year and the results of upcoming year. If I keep doing this in the upcoming years, I can compare all the results and see if there is a specific movement. Are we growing to each other? Or do we keep our own taste in music? Or maybe our musictastes will melt together. I think it would be interesting to keep looking at this.